For Pachi the data has already been posted by Jean-Loup, and for Fuego there
is also a published ceiling according to the scaling study in the (free
download) paper from Richard Segal on BlueFuego (someone sent out a link
within the last two weeks, too lazy to check who). Being limited to Linux
rules my cpu's out, otherwise I wouldn't mind donating some cores. Also, if
David is willing to donate executables to the study, I don't see why that's
not acceptable, i.e., that's as good as "free" to me.

I am running a mini-scaling experiment on CGOS 9x9 right now using the AMAF
standard bots (as originally implemented by Don Dailey) with and without
edge-move veto's. They now have all received about 10,000 games, I should
stop it and summarize the data (this experiment also confirmed that
Elo-ratings converged to below 9 Elo sigma). Each level doubles the number
of playouts. Obviously, CGOS time settings won't allow too many doublings
before you run out of time, so there is only 6 levels for each
configuration. I did find it noteworthy that the ratings in the Bayes-Elo
list are more as expected (in that the ranking is what I would expect) as
compared to the incremental ones on the server itself.

René

On Sat, Jun 18, 2011 at 8:54 PM, David Fotland <[email protected]>wrote:

> Maybe Pachi and Fuego?  They are both free I think.****
>
> ** **
>
> David****
>
> ** **
>
> *From:* [email protected] [mailto:
> [email protected]] *On Behalf Of *Don Dailey
> *Sent:* Saturday, June 18, 2011 3:11 PM
>
> *To:* [email protected]
> *Subject:* Re: [Computer-go] scalability study****
>
> ** **
>
> ** **
>
> Unfortunatelly again, all of my pcs are busy for my research
>
> work/experiments in this year or more (perhaps, until I'll finish my
> thesis).****
>
> ** **
>
> If we do a study it won't be dependent on any particular computer,  it will
> be a big group effort and we should only use programs that are available on
> all the target machines.   ****
>
> ** **
>
> It will be very easy to manage if it's done via ssh to remote linux
> computers.  It's also possible to set up remote ssh on windows,  but I have
> no idea how to do that and don't want to get into it unless someone else is
> willing to take the lead on this.   Stuff like this is always hard in
> Windows and easy on Unix. ****
>
> ** **
>
> What are the best 2 programs that are freely available for such a study
> (that we do not have to pay for) so that we can get some people willing to
> give me an account on their machine?    ****
>
> ** **
>
> Don****
>
> ** **
>
> ** **
>
> _______________________________________________
> Computer-go mailing list
> [email protected]
> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
>
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go

Reply via email to